Initial release: PickUP V1 LeRobotDataset v3.0 public sample (6 episodes)

README.md

@@ -0,0 +1,190 @@
+---
+license: other
+license_name: to-be-confirmed
+task_categories:
+- robotics
+tags:
+- lerobot
+- manipulation
+- unity-simulation
+- pick-and-place
+- so-100
+- imitation-learning
+size_categories:
+- n<1K
+configs:
+- config_name: default
+  data_files:
+  - split: train
+    path: data/chunk-000/file-000.parquet
+---
+
+# PickUP V1 — Unity Robot Pick & Place Dataset
+
+> **⚠️ Simulation Data — Not Physical Robot Data**  
+> This dataset was generated and validated in a high-fidelity Unity physics simulation environment (`PickUP Simulator V1.0`) utilizing an ArticulationBody-based 6-DoF robotic manipulator.
+
+---
+
+## 1. Overview
+**PickUP V1** is an open demonstration sample dataset formatted in official **LeRobotDataset v3.0** standard. It contains 6 verified expert demonstration episodes of single-arm cube pick-and-place manipulation across 6 spatial spawn zones.
+
+- **Dataset Standard**: `LeRobotDataset v3.0`
+- **Total Episodes**: `6`
+- **Total Frames**: `1,266` (211 frames/episode at 30 FPS, duration: 7.03s/episode)
+- **Task Success Rate**: `100%` (6/6 validated successful trajectories)
+- **Cameras**: Fixed Top-Down Camera (`observation.images.top`) and Wrist Eye-in-Hand Camera (`observation.images.wrist`)
+
+---
+
+## 2. Demo
+A 21-second demonstration video showcasing multi-angle synchronized execution across representative workspace zones is available in the [`demo/`](./demo/) directory:
+- **Demo Video Path**: [`demo/pickup_v1_demo.mp4`](./demo/pickup_v1_demo.mp4)
+- **Format**: Synchronized Side-by-Side (Top-Down View + Wrist Eye-in-Hand View) with phase annotations.
+
+---
+
+## 3. Task Definition
+- **Task Instruction**: `"Pick up the red cube and place it in the green tray."`
+- **Target Object**: Red cubic block (dimensions: $35\text{ mm} \times 35\text{ mm} \times 35\text{ mm}$, mass: $100\text{ g}$).
+- **Target Placement Zone**: Green receptacle tray located at $[0.180, 0.370, 0.220]\text{ m}$.
+- **Motion Sequence**:
+  1. **Phase 1 (Approach)**: End-effector moves from home pose to pre-grasp alignment above target cube.
+  2. **Phase 2 (Grasp)**: Parallel fingers close with validated bilateral contact.
+  3. **Phase 3 (Vertical Lift)**: Cube is lifted $+50\text{ mm}$ vertically to ensure ground clearance.
+  4. **Phase 4 (Transfer)**: 3D spatial interpolation toward the target receptacle tray.
+  5. **Phase 5 (Place & Release)**: Cube is lowered into the tray and gripper releases.
+  6. **Phase 6 (Home Return)**: Arm safely returns to ready pose.
+
+---
+
+## 4. Dataset Summary
+
+| Parameter | Value |
+|---|---|
+| **Codebase Version** | `v3.0` |
+| **Robot Architecture** | SO-100 inspired 6-DoF single arm with parallel gripper |
+| **Total Episodes** | `6` |
+| **Total Frames** | `1,266` |
+| **Frames per Episode** | `211` |
+| **Frame Rate** | `30 FPS` |
+| **Episode Duration** | `7.03 seconds` |
+| **Observation Streams** | 2 Video Streams ($240 \times 180$) + 6-DoF Joint State Tensor |
+| **Action Space** | 6-DoF Target Joint Positions Tensor |
+
+---
+
+## 5. Dataset Structure
+
+```text
+PickUP_HuggingFace_Public/
+├── README.md                 # Hugging Face Dataset Card
+├── data/
+│   └── chunk-000/
+│       └── file-000.parquet  # 1,266 data frames (state, action, timestamps, indices)
+├── meta/
+│   ├── info.json             # Dataset specifications and feature schemas
+│   ├── stats.json            # Recalculated channel-wise and feature statistics
+│   ├── tasks.parquet         # Task index 0 mapping
+│   ├── episode_outcomes.parquet # Episode success and metadata records
+│   ├── provenance.json       # Source episode provenance mapping
+│   └── episodes/
+│       └── chunk-000/
+│           └── file-000.parquet # Per-episode indices, timestamps, and statistics
+├── videos/
+│   ├── observation.images.top/
+│   │   └── chunk-000/
+│   │       └── file-000.mp4  # Top-down camera stream (240x180, H.264, 30 FPS)
+│   └── observation.images.wrist/
+│       └── chunk-000/
+│           └── file-000.mp4  # Wrist eye-in-hand stream (240x180, H.264, 30 FPS)
+├── demo/
+│   └── pickup_v1_demo.mp4    # 21-second demonstration video
+└── docs/
+    ├── DATASET_VALIDATION.md # Detailed validation checklist and test results
+    └── V1_LIMITATIONS_SUMMARY.md # Hardware envelope and known limitation boundaries
+```
+
+---
+
+## 6. Observation & Action Features
+
+| Feature Key | Type | Shape / Format | Description |
+|---|---|:---:|---|
+| **`observation.images.top`** | Video | `(180, 240, 3)` | RGB Top-Down fixed overview camera ($60^\circ$ FOV, 30 FPS) |
+| **`observation.images.wrist`** | Video | `(180, 240, 3)` | RGB Eye-in-Hand camera mounted on J4 wrist flex ($75^\circ$ FOV, 30 FPS) |
+| **`observation.state`** | Float32 Tensor | `(6,)` | Joint angles: `[shoulder_pan, shoulder_lift, elbow_flex, wrist_flex, wrist_roll, gripper]` |
+| **`action`** | Float32 Tensor | `(6,)` | Target joint angles: `[shoulder_pan, shoulder_lift, elbow_flex, wrist_flex, wrist_roll, gripper]` |
+| **`timestamp`** | Float32 | `(1,)` | Frame timestamp in seconds relative to episode start ($0.0\text{s} \sim 7.0\text{s}$) |
+| **`frame_index`** | Int64 | `(1,)` | Episode-relative frame counter ($0 \sim 210$) |
+| **`episode_index`** | Int64 | `(1,)` | Episode index ($0 \sim 5$) |
+| **`index`** | Int64 | `(1,)` | Global dataset sample index ($0 \sim 1,265$) |
+| **`task_index`** | Int64 | `(1,)` | Task index (`0`) |
+
+---
+
+## 7. Spawn Zone Coverage
+
+The 6 sample episodes uniformly cover all quadrants of the reachable workspace:
+
+| Sample Ep | Spawn Zone | Cube Initial Position (m) | Random Seed | Placement Error | Status |
+|:---:|:---:|:---:|:---:|:---:|:---:|
+| **0** | `left_front` | `[-0.135, 0.389, 0.275]` | 20260826 | $< 1.5\text{ mm}$ | **SUCCESS** |
+| **1** | `left_back` | `[-0.148, 0.389, 0.186]` | 20260827 | $< 1.5\text{ mm}$ | **SUCCESS** |
+| **2** | `center_front` | `[-0.016, 0.389, 0.287]` | 20260828 | $< 2.0\text{ mm}$ | **SUCCESS** |
+| **3** | `center_back` | `[-0.025, 0.389, 0.174]` | 20260829 | $< 2.0\text{ mm}$ | **SUCCESS** |
+| **4** | `right_front` | `[0.066, 0.389, 0.276]` | 20260830 | $< 1.5\text{ mm}$ | **SUCCESS** |
+| **5** | `right_back` | `[0.053, 0.389, 0.185]` | 20260831 | $< 1.5\text{ mm}$ | **SUCCESS** |
+
+---
+
+## 8. Validation & Quality Assurance
+- **LeRobotDataset v3.0 Compatibility**: 100% compliant with Hugging Face LeRobot loader.
+- **Zero Frame Errors**: Full sequential traversal of 1,266 frames verified without decode drops.
+- **Detailed Report**: See [`docs/DATASET_VALIDATION.md`](./docs/DATASET_VALIDATION.md) for the complete QA checklist.
+
+---
+
+## 9. Intended Uses
+- **Benchmarking Imitation Learning Policies**: ACT (Action Chunking with Transformers), Diffusion Policy, and VQ-BET.
+- **Multimodal Visual-Motor Policy Training**: Evaluation of multi-camera (Top + Wrist) sensor fusion.
+- **Sim-to-Sim Policy Transfer**: Transferability evaluation across robotic simulation engines.
+
+---
+
+## 10. Limitations & Known Boundaries
+- **Simulation Domain**: Generated in Unity simulation; real-world sim-to-real transfer requires domain randomization.
+- **Payload & Workspace Envelope**: Standard V1 parallel gripper is validated for $35\text{mm}$ cubes up to $100\text{g}$ within radius $r \in [0.16\text{m}, 0.22\text{m}]$.
+- **Detailed Limitations**: Refer to [`docs/V1_LIMITATIONS_SUMMARY.md`](./docs/V1_LIMITATIONS_SUMMARY.md) for empirical hardware and stress-testing boundary records.
+
+---
+
+## 11. Simulator Information
+The dataset was generated using the **PickUP V1 Unity simulation environment** (`PickUP Simulator V1.0`) built with Unity 6000.3.22f1 and PhysX ArticulationBody kinematics. Simulator source code and master archives are maintained separately and are not included in this dataset package.
+
+---
+
+## 12. License
+**License: Apache 2.0 (Candidate)**  
+*All visual assets in the dataset videos were audited and confirmed to be 100% generated from Unity built-in primitives and procedural materials with zero third-party copyright dependencies. See [`docs/ASSET_LICENSE_AUDIT.md`](./docs/ASSET_LICENSE_AUDIT.md) for the complete 1:1 audit matrix.*
+
+---
+
+## 13. Citation
+If you use this dataset in your research or benchmarks, please cite:
+
+```bibtex
+@dataset{pickup_v1_dataset_2026,
+  title={PickUP V1: Unity Robot Pick & Place Dataset},
+  author={PickUP Project Contributors},
+  year={2026},
+  publisher={Hugging Face},
+  howpublished={\url{https://huggingface.co/datasets/devorah-ai-2016/PickUP-PickPlace-V1}}
+}
+```
+
+---
+
+## 14. Contact & Project Information
+- **Project**: PickUP Robotics Manipulation Initiative
+- **Documentation**: See [`docs/`](./docs/) for validation and limitations documents.

data/chunk-000/file-000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f768002698b96e5d0be489ded7c80a956999424d6e7010840f201712b87de36d
+size 46897

demo/pickup_v1_demo.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fedcf2b25a29fbdc48083134cfd0a7de14c51044d8df0700a28200b678e4cbb3
+size 925259

docs/ASSET_LICENSE_AUDIT.md

@@ -0,0 +1,59 @@
+# PickUP V1 시각 에셋 라이선스 및 지식재산권(IP) 감사 보고서 (ASSET_LICENSE_AUDIT)
+
+---
+
+## 1. 감사 개요 및 목적
+본 감사는 Hugging Face 공개용 데이터셋(`PickUP_HuggingFace_Public`) 영상(`observation.images.top`, `observation.images.wrist`) 및 데모 비디오(`pickup_v1_demo.mp4`)에 **실제로 노출·렌더링되는 모든 시각 에셋(Visual Assets)**의 원본 파일명, 제작 방식, 출처, 원 라이선스 및 영상 재배포 허용 여부를 1:1로 전수 조사한 공식 감사 보고서입니다.
+
+- **감사 기준**: 직접 제작 / AI 생성 / 외부 다운로드를 엄격히 구분하고, 객관적 증빙이 없는 항목은 **`UNKNOWN`**으로 처리하며 임의 추정을 일체 배제함.
+- **Hugging Face 업로드**: 미수행 (로컬 라이선스 감사만 실시).
+
+---
+
+## 2. 데이터셋 렌더링 씬 에셋 1:1 전수 감사 매트릭스
+
+데이터셋 생성의 기준 씬인 `SMB_Assist_Robot_V1_CORE_BASELINE_LOCKED_20260827.unity`의 씬 파일 YAML 및 프로젝트 에셋을 전수 분석한 1:1 대조 결과는 다음과 같습니다.
+
+| 에셋명 (Asset Name) | 영상 노출 역할 | 원본 파일 / 코드 경로 | 제작 방식 (Creation Method) | 출처 구분 | 원 라이선스 (Original License) | 재배포 및 영상 공개 허용 여부 | 감사 판정 |
+|---|---|---|---|:---:|---|:---:|:---:|
+| **Robot V1 메인 바디 (`Base_Mount`, `J1`~`J5`)** | 6-DoF 로봇 매니퓰레이터 본체 | `Assets/Editor/SMBRobotArmBuilder.cs` | Unity 빌트인 프리미티브(`PrimitiveType.Cylinder`, `Cube`) C# 절차적 기하구조 조립 | **직접 제작 (In-House Procedural)** | Unity Engine Built-in Primitives (Unity EULA) | **허용 (PERMITTED)**<br>엔진 기본 도형 렌더링 결과물은 제작자 완전 귀속 | **`CONFIRMED`** |
+| **Standard Gripper (`TestGripper_V1`)** | 로봇 엔드이펙터 좌/우 평행 핑거 | `Assets/Editor/SMBRobotArmBuilder.cs` | Unity 빌트인 프리미티브(`PrimitiveType.Cube`) C# 절차적 기하구조 조립 | **직접 제작 (In-House Procedural)** | Unity Engine Built-in Primitives (Unity EULA) | **허용 (PERMITTED)** | **`CONFIRMED`** |
+| **Heavy Box Gripper (`HeavyBoxGripper_V1`)** | 중량물 전용 핑거 (1.8mm 립) | `Assets/Editor/SMBHeavyBoxRobotBuilder.cs` | Unity 빌트인 프리미티브(`PrimitiveType.Cube`) C# 절차적 기하구조 조립 | **직접 제작 (In-House Procedural)** | Unity Engine Built-in Primitives (Unity EULA) | **허용 (PERMITTED)** | **`CONFIRMED`** |
+| **Target Cube (`Target_Cube_V1`)** | 파지 대상 35mm 붉은색 큐브 | `Scenes/SMB_Assist_Robot_V1_...unity` | Unity 빌트인 프리미티브(`PrimitiveType.Cube`) | **직접 제작 (In-House Primitive)** | Unity Engine Built-in Primitives (Unity EULA) | **허용 (PERMITTED)** | **`CONFIRMED`** |
+| **Placement Tray (`Cube_Target_Zone_V1`)** | 목표 안착 트레이 (에메랄드/황색) | `Scenes/SMB_Assist_Robot_V1_...unity` | Unity 빌트인 프리미티브(`PrimitiveType.Cube`) | **직접 제작 (In-House Primitive)** | Unity Engine Built-in Primitives (Unity EULA) | **허용 (PERMITTED)** | **`CONFIRMED`** |
+| **Table & Ground Plane** | 작업대 및 바닥 평면 | `Scenes/SMB_Assist_Robot_V1_...unity` | Unity 빌트인 프리미티브(`PrimitiveType.Cube`, `Plane`) | **직접 제작 (In-House Primitive)** | Unity Engine Built-in Primitives (Unity EULA) | **허용 (PERMITTED)** | **`CONFIRMED`** |
+| **Materials (`M_RobotArm_*`, `M_TargetCube_*`, `M_Table` 등)** | 로봇/물체 표면 재질 및 색상 | `Assets/Materials/*.mat` | URP Lit 셰이더 기반 단색 RGB(`_BaseColor`) 지정 머티리얼 (외부 텍스처 0개) | **직접 제작 (In-House Procedural)** | Unity Universal Render Pipeline (Unity Companion License) | **허용 (PERMITTED)** | **`CONFIRMED`** |
+| **Shader (`Universal Render Pipeline/Lit`)** | 표면 조명 연산 셰이더 | Unity 공식 URP 패키지 (`com.unity.render-pipelines.universal`) | Unity 공식 렌더 파이프라인 빌트인 셰이더 | **엔진 공식 패키지** | Unity Companion License / Unity EULA | **허용 (PERMITTED)**<br>렌더링 픽셀 출력물의 상업/비상업 공개 자유 | **`CONFIRMED`** |
+| **Textures / Bitmap Maps** | 알베도, 노멀, 금속성 맵 등 이미지 파일 | 사용되지 않음 (머티리얼 텍스처 슬롯 비어있음) | 순수 단색 상수 컬러(Procedural Solid Color) 연산 | **미사용 (None)** | 해당 없음 (N/A) | **허용 (PERMITTED)**<br>서드파티 이미지 저작권 의존성 0건 | **`CONFIRMED`** |
+| **Background / Skybox** | 카메라 배경 | 카메라 `SolidColor` 단색 배경 (`RGB: 0.92, 0.92, 0.93`) | Unity Camera 배경 단색 클리어 설정 | **직접 설정 (In-House)** | 해당 없음 (N/A) | **허용 (PERMITTED)**<br>외부 HDRI/스카이박스 에셋 의존성 0건 | **`CONFIRMED`** |
+| **Legacy Models (`Assets/Models/*.obj`)** | 초기 프로토타입 참고용 OBJ (Part 0~7) | `Assets/Models/Part_*.obj` | Wavefront OBJ 3D 메쉬 파일 (초기 탐색 단계) | **초기 참조본 (데이터셋 씬 미참조·미렌더링)** | 출처 메타데이터 미기재 | **데이터셋 영상에 일체 미포함**<br>(소스 아카이브 배포 시: **`UNKNOWN`**) | **`UNKNOWN` (Unused)** |
+
+---
+
+## 3. 핵심 조사 및 검증 결과
+
+### A. 데이터셋 영상의 100% 순수 빌트인 프리미티브 기반 렌더링 확인
+- Locked Baseline Scene(`SMB_Assist_Robot_V1_CORE_BASELINE_LOCKED_20260827.unity`) YAML에 등록된 모든 3D 메쉬 GUID를 전수 조사한 결과, `Assets/Models/*.obj`를 포함한 **어떠한 외부 3D 모델(FBX, OBJ, GLTF 등)도 참조되거나 렌더링되지 않음(0건)**이 확인되었습니다.
+- 영상에 등장하는 모든 시각 요소는 Unity 엔진의 표준 기하학적 기본도형(`PrimitiveType.Cylinder`, `PrimitiveType.Cube`, `PrimitiveType.Plane`)과 순수 RGB 컬러 머티리얼로만 절차적 렌더링되었습니다.
+
+### B. 서드파티 텍스처 및 에셋스토어 저작권 의존성 0건
+- 프로젝트 내 머티리얼 15종을 전수 분석한 결과, 외부 이미지 파일(PNG, JPG, TGA 등)을 텍스처 맵으로 참조하는 머티리얼은 **0건**입니다.
+- 배경 화면 역시 외부 HDRI나 스카이박스 텍스처를 사용하지 않고 순수 `SolidColor` 단색으로 처리되어 제3자 저작권 침해 요소가 전혀 없습니다.
+
+### C. 미사용 레거시 에셋(`Assets/Models/*.obj`)의 UNKNOWN 분리 처리
+- `Assets/Models/` 경로에 존재하는 9개의 Wavefront OBJ 파일은 초기 연구 탐색 시 사용된 참조 파일로서, 원 저작자 및 라이선스 증빙이 프로젝트 내 메타데이터에 명시되어 있지 않습니다.
+- 따라서 원칙에 따라 본 파일들의 원 라이선스는 **`UNKNOWN`**으로 엄격히 분류하며, **공개 데이터셋 영상에는 이 파일들이 일체 포함되지 않았음**을 공식 확인합니다.
+
+---
+
+## 4. 최종 공개 라이선스 권고안
+
+1. **데이터셋 자체 (`PickUP_HuggingFace_Public`)**:
+   - 영상 및 Parquet 데이터는 순수 Unity 빌트인 프리미티브의 렌더링 결과물이므로, 제3자 지식재산권 침해 우려 없이 **`Apache-2.0`** 또는 **`CC-BY-4.0` / `MIT`** 등 표준 오픈소스 라이선스로 공개가 가능합니다.
+2. **시뮬레이터 마스터 소스 (`PickUP_Simulator_Master`)**:
+   - 시뮬레이터 소스코드 전체를 외부 공개할 경우에는 `Assets/Models/*.obj` 레거시 파일을 제외(Sanitize)하거나 라이선스를 확정한 후 배포할 것을 권고합니다.
+
+---
+
+> **결론 및 판정: DATASET VIDEO ASSET AUDIT 100% CLEAR (CONFIRMED)**  
+> 공개 대상 데이터셋 비디오 및 데이터는 서드파티 저작권 의존성 없이 안전하게 공개 가능한 상태입니다.

docs/DATASET_VALIDATION.md

@@ -0,0 +1,47 @@
+# PickUP V1 Dataset Validation Summary
+
+This document summarizes the official data integrity and consistency verification performed on the **PickUP V1 Public Sample Dataset**.
+
+---
+
+## 1. Summary of Verified Scope
+- **Dataset Format**: `LeRobotDataset v3.0`
+- **Total Selected Episodes**: `6`
+- **Total Frames**: `1,266` (211 frames per episode at 30 FPS, duration: 7.03s/episode)
+- **Cameras**:
+  - `observation.images.top` (240x180 RGB video stream, 30 FPS)
+  - `observation.images.wrist` (240x180 RGB video stream, 30 FPS)
+- **Task Description**: `"Pick up the red cube and place it in the green tray."`
+- **Robot Type**: `so-100` inspired 6-DoF manipulator in Unity simulation
+
+---
+
+## 2. Validation Checklist Results
+
+| Test Category | Description | Result |
+|---|---|:---:|
+| **LeRobot Loader Initialization** | Successfully instantiated with official `LeRobotDataset(root=..., repo_id=...)` class | **PASS** |
+| **Full Frame Sequential Access** | All 1,266 frames iterated and decoded sequentially with 0 read errors | **PASS** |
+| **Observation Tensor Shapes** | `observation.images.top`: `(3, 180, 240)`<br>`observation.images.wrist`: `(3, 180, 240)`<br>`observation.state`: `(6,)`<br>`action`: `(6,)` | **PASS** |
+| **Index & Timestamp Consistency** | Continuous indexing (`index: 0..1265`), episode-relative indexing (`frame_index: 0..210`), and timestamps (`0.0s .. 7.0s`) perfectly aligned | **PASS** |
+| **Metadata & Statistics Alignment** | `stats.json` recalculated specifically for the 6 public episodes (min, max, mean, std, quantiles) | **PASS** |
+| **Episode Outcomes Alignment** | `episode_outcomes.parquet` contains 6 verified successful episode records (`is_success: True`, `reward: 1.0`) | **PASS** |
+| **Non-Public Data Leakage Test** | Verified zero leakage of the remaining 54 non-public master episodes from parquets, videos, and metadata | **PASS** |
+
+---
+
+## 3. Verified Spawn Zone Representation
+
+| Sample Ep Index | Spawn Zone | Initial Cube Position (m) | Tray Position (m) | Outcome |
+|:---:|:---:|:---:|:---:|:---:|
+| **0** | `left_front` | `[-0.135, 0.389, 0.275]` | `[0.180, 0.370, 0.220]` | **SUCCESS (1.0)** |
+| **1** | `left_back` | `[-0.148, 0.389, 0.186]` | `[0.180, 0.370, 0.220]` | **SUCCESS (1.0)** |
+| **2** | `center_front` | `[-0.016, 0.389, 0.287]` | `[0.180, 0.370, 0.220]` | **SUCCESS (1.0)** |
+| **3** | `center_back` | `[-0.025, 0.389, 0.174]` | `[0.180, 0.370, 0.220]` | **SUCCESS (1.0)** |
+| **4** | `right_front` | `[0.066, 0.389, 0.276]` | `[0.180, 0.370, 0.220]` | **SUCCESS (1.0)** |
+| **5** | `right_back` | `[0.053, 0.389, 0.185]` | `[0.180, 0.370, 0.220]` | **SUCCESS (1.0)** |
+
+---
+
+> **Verification Status: 100% PASS**  
+> This dataset has been confirmed ready for public evaluation and imitation learning benchmarks.

docs/V1_LIMITATIONS_SUMMARY.md

@@ -0,0 +1,40 @@
+# PickUP V1 Known Limitations Summary
+
+This document outlines the confirmed operational and hardware boundaries of the **PickUP V1** robot system, established through empirical testing across Tasks 1–6.
+
+---
+
+## 1. Validated Safe Operating Envelope
+The PickUP V1 dataset operates within the following confirmed safe operating domain:
+- **Target Object**: 35mm cubic geometry (mass: 50g ~ 100g).
+- **Workspace Radius**: Radial distance $r \in [0.16\text{m}, 0.22\text{m}]$ from base origin.
+- **Task Success Rate**: 100% across the 6 primary spawn zones.
+- **Mean Placement Accuracy**: $1.83\text{ mm}$ (maximum error: $2.74\text{ mm}$).
+
+---
+
+## 2. Confirmed Boundary Limitations
+
+### A. Operational Payload Envelope (`V1-LIM-001`, `V1-LIM-002`)
+- **Standard Parallel Head (`TestGripper_V1`)**: Reliably handles up to $100\text{g}$. At payloads $\ge 250\text{g}$, purely planar rubber pad friction is exceeded during vertical lift acceleration, causing object slip/drop.
+- **Heavy Box Head (`HeavyBoxGripper_V1`)**: Achieves 100% grasp and transport up to $400\text{g}$ using 25N clamping force and a 1.8mm geometric under-lip.
+- **Shoulder Torque Saturation Threshold**: At payloads $\ge 500\text{g}$ with full extension ($r \approx 0.20\text{m}$), gravitational moment arm causes J2 shoulder joint torque demand to reach $1.097\text{ N}\cdot\text{m}$ at 500g and exceed the experimental limit of **`1.287 N·m`** at 600g, causing actuator saturation and trajectory sag.
+
+### B. Reachability & Workspace Boundaries (`V1-LIM-003`)
+- **Near Boundary**: Target positions closer than $0.16\text{m}$ ($r < 0.16\text{m}$) cannot be reached while maintaining a top-down vertical approach angle due to kinematic joint limits (J2/J3).
+- **Far Boundary**: Target positions farther than $0.235\text{m}$ ($r > 0.235\text{m}$) exceed total link reach geometry.
+
+### C. Object Dimension Grasp Limits (`V1-LIM-004`)
+- **Small Objects (25mm)**: Insufficient pad contact area at finger tips leads to unstable retention.
+- **Large Objects (45mm)**: Causes asymmetric contact and grasp center offset. The standard gripper is optimized for $35\text{mm}$ targets.
+
+### D. Wrist Compliance under Eccentric Cantilever Load (`V1-LIM-005`)
+- When handling off-center or wide objects ($>50\text{mm}$ offset), single-shear compact wrist servos experience torsional compliance, increasing placement error.
+
+### E. Dynamic Transfer Acceleration Limits (`V1-LIM-006`)
+- Horizontal transfer speeds exceeding $0.15\text{ m/s}$ during open-loop kinematic interpolation induce inertial shear forces that risk micro-slip on unconstrained payloads $>100\text{g}$.
+
+---
+
+## 3. Dataset Integrity Note
+The limitations described above represent physical/hardware boundaries tested during stress-testing phases. **They do not affect the validity of this public dataset**, which is entirely contained within the 100% verified safe operational envelope ($100\text{g}$ payload, standard workspace zones).

meta/episode_outcomes.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a680546d4e3cec9e7c91d267f0d2902a82f948a64765536d43bba8ce248124e8
+size 13785

meta/episodes/chunk-000/file-000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:09273974bead74e59ff11651245d43edc3175e79702a0b7700def850b8cb771c
+size 94878

meta/info.json

@@ -0,0 +1,127 @@
+{
+  "codebase_version": "v3.0",
+  "robot_type": "so-100",
+  "total_episodes": 6,
+  "total_frames": 1266,
+  "total_tasks": 1,
+  "chunks_size": 1000,
+  "data_files_size_in_mb": 100,
+  "video_files_size_in_mb": 200,
+  "fps": 30,
+  "splits": {
+    "train": "0:6"
+  },
+  "data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
+  "video_path": "videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4",
+  "features": {
+    "observation.state": {
+      "dtype": "float32",
+      "shape": [
+        6
+      ],
+      "names": [
+        "shoulder_pan",
+        "shoulder_lift",
+        "elbow_flex",
+        "wrist_flex",
+        "wrist_roll",
+        "gripper"
+      ]
+    },
+    "action": {
+      "dtype": "float32",
+      "shape": [
+        6
+      ],
+      "names": [
+        "shoulder_pan",
+        "shoulder_lift",
+        "elbow_flex",
+        "wrist_flex",
+        "wrist_roll",
+        "gripper"
+      ]
+    },
+    "observation.images.top": {
+      "dtype": "video",
+      "shape": [
+        180,
+        240,
+        3
+      ],
+      "names": [
+        "height",
+        "width",
+        "channels"
+      ],
+      "info": {
+        "video.height": 180,
+        "video.width": 240,
+        "video.codec": "h264",
+        "video.pix_fmt": "yuv420p",
+        "video.is_depth_map": false,
+        "video.fps": 30,
+        "video.channels": 3,
+        "has_audio": false
+      }
+    },
+    "observation.images.wrist": {
+      "dtype": "video",
+      "shape": [
+        180,
+        240,
+        3
+      ],
+      "names": [
+        "height",
+        "width",
+        "channels"
+      ],
+      "info": {
+        "video.height": 180,
+        "video.width": 240,
+        "video.codec": "h264",
+        "video.pix_fmt": "yuv420p",
+        "video.is_depth_map": false,
+        "video.fps": 30,
+        "video.channels": 3,
+        "has_audio": false
+      }
+    },
+    "timestamp": {
+      "dtype": "float32",
+      "shape": [
+        1
+      ],
+      "names": null
+    },
+    "frame_index": {
+      "dtype": "int64",
+      "shape": [
+        1
+      ],
+      "names": null
+    },
+    "episode_index": {
+      "dtype": "int64",
+      "shape": [
+        1
+      ],
+      "names": null
+    },
+    "index": {
+      "dtype": "int64",
+      "shape": [
+        1
+      ],
+      "names": null
+    },
+    "task_index": {
+      "dtype": "int64",
+      "shape": [
+        1
+      ],
+      "names": null
+    }
+  }
+}

meta/provenance.json

@@ -0,0 +1,94 @@
+{
+  "source_master_dataset": "lerobot-sim-cube-pick-place-expert-success-v0.3",
+  "total_master_episodes": 60,
+  "total_sample_episodes": 6,
+  "fps": 30,
+  "frames_per_episode": 211,
+  "total_frames": 1266,
+  "episodes_mapping": [
+    {
+      "sample_episode_index": 0,
+      "master_episode_index": 0,
+      "episode_id": "ep_000_left_front",
+      "spawn_zone": "left_front",
+      "cube_initial_position": [
+        -0.13500000536441803,
+        0.3889999985694885,
+        0.2750000059604645
+      ],
+      "random_seed": 20260826,
+      "is_success": true,
+      "reward": 1.0
+    },
+    {
+      "sample_episode_index": 1,
+      "master_episode_index": 10,
+      "episode_id": "ep_010_left_back",
+      "spawn_zone": "left_back",
+      "cube_initial_position": [
+        -0.14800000190734863,
+        0.3889999985694885,
+        0.1860000044107437
+      ],
+      "random_seed": 20260827,
+      "is_success": true,
+      "reward": 1.0
+    },
+    {
+      "sample_episode_index": 2,
+      "master_episode_index": 20,
+      "episode_id": "ep_020_center_front",
+      "spawn_zone": "center_front",
+      "cube_initial_position": [
+        -0.01600000075995922,
+        0.3889999985694885,
+        0.28700000047683716
+      ],
+      "random_seed": 20260828,
+      "is_success": true,
+      "reward": 1.0
+    },
+    {
+      "sample_episode_index": 3,
+      "master_episode_index": 30,
+      "episode_id": "ep_030_center_back",
+      "spawn_zone": "center_back",
+      "cube_initial_position": [
+        -0.02500000037252903,
+        0.3889999985694885,
+        0.17399999499320984
+      ],
+      "random_seed": 20260829,
+      "is_success": true,
+      "reward": 1.0
+    },
+    {
+      "sample_episode_index": 4,
+      "master_episode_index": 40,
+      "episode_id": "ep_040_right_front",
+      "spawn_zone": "right_front",
+      "cube_initial_position": [
+        0.06599999964237213,
+        0.3889999985694885,
+        0.2759999930858612
+      ],
+      "random_seed": 20260830,
+      "is_success": true,
+      "reward": 1.0
+    },
+    {
+      "sample_episode_index": 5,
+      "master_episode_index": 50,
+      "episode_id": "ep_050_right_back",
+      "spawn_zone": "right_back",
+      "cube_initial_position": [
+        0.05299999937415123,
+        0.3889999985694885,
+        0.1850000023841858
+      ],
+      "random_seed": 20260831,
+      "is_success": true,
+      "reward": 1.0
+    }
+  ]
+}

meta/stats.json

@@ -0,0 +1,642 @@
+{
+  "observation.state": {
+    "min": [
+      0.0,
+      -0.3499999940395355,
+      0.75,
+      0.4000000059604645,
+      0.0,
+      0.0
+    ],
+    "max": [
+      2.256525754928589,
+      0.556190550327301,
+      2.066716194152832,
+      1.791893482208252,
+      0.0,
+      0.949999988079071
+    ],
+    "mean": [
+      1.6155105829238892,
+      0.2580343186855316,
+      1.2688515186309814,
+      1.3372653722763062,
+      0.0,
+      0.4322294294834137
+    ],
+    "std": [
+      0.6266899108886719,
+      0.24129623174667358,
+      0.31968557834625244,
+      0.30681946873664856,
+      0.0,
+      0.4452095627784729
+    ],
+    "count": [
+      1266,
+      1266,
+      1266,
+      1266,
+      1266,
+      1266
+    ],
+    "q01": [
+      0.043288759887218475,
+      -0.3378557860851288,
+      0.7576088309288025,
+      0.4458378553390503,
+      0.0,
+      0.0
+    ],
+    "q10": [
+      0.7039938569068909,
+      -0.10897818207740784,
+      0.8641321063041687,
+      0.9219600558280945,
+      0.0,
+      0.0
+    ],
+    "q50": [
+      1.8055189847946167,
+      0.3991503417491913,
+      1.249741554260254,
+      1.3605610132217407,
+      0.0,
+      0.21111111342906952
+    ],
+    "q90": [
+      2.256525754928589,
+      0.4581843614578247,
+      1.7559632062911987,
+      1.6821633577346802,
+      0.0,
+      0.949999988079071
+    ],
+    "q99": [
+      2.256525754928589,
+      0.556190550327301,
+      2.066716194152832,
+      1.745741367340088,
+      0.0,
+      0.949999988079071
+    ]
+  },
+  "action": {
+    "min": [
+      0.0,
+      -0.3499999940395355,
+      0.75,
+      0.4000000059604645,
+      0.0,
+      0.0
+    ],
+    "max": [
+      2.256525754928589,
+      0.556190550327301,
+      2.066716194152832,
+      1.791893482208252,
+      0.0,
+      0.949999988079071
+    ],
+    "mean": [
+      1.6155105829238892,
+      0.2580343186855316,
+      1.2688515186309814,
+      1.3372653722763062,
+      0.0,
+      0.4322294294834137
+    ],
+    "std": [
+      0.6266899108886719,
+      0.24129623174667358,
+      0.31968557834625244,
+      0.30681946873664856,
+      0.0,
+      0.4452095627784729
+    ],
+    "count": [
+      1266,
+      1266,
+      1266,
+      1266,
+      1266,
+      1266
+    ],
+    "q01": [
+      0.043288759887218475,
+      -0.3378557860851288,
+      0.7576088309288025,
+      0.4458378553390503,
+      0.0,
+      0.0
+    ],
+    "q10": [
+      0.7039938569068909,
+      -0.10897818207740784,
+      0.8641321063041687,
+      0.9219600558280945,
+      0.0,
+      0.0
+    ],
+    "q50": [
+      1.8055189847946167,
+      0.3991503417491913,
+      1.249741554260254,
+      1.3605610132217407,
+      0.0,
+      0.21111111342906952
+    ],
+    "q90": [
+      2.256525754928589,
+      0.4581843614578247,
+      1.7559632062911987,
+      1.6821633577346802,
+      0.0,
+      0.949999988079071
+    ],
+    "q99": [
+      2.256525754928589,
+      0.556190550327301,
+      2.066716194152832,
+      1.745741367340088,
+      0.0,
+      0.949999988079071
+    ]
+  },
+  "timestamp": {
+    "min": [
+      0.0
+    ],
+    "max": [
+      7.0
+    ],
+    "mean": [
+      3.5000000003707634
+    ],
+    "std": [
+      2.030325645972856
+    ],
+    "count": [
+      1266
+    ],
+    "q01": [
+      0.06666667014360428
+    ],
+    "q10": [
+      0.699999988079071
+    ],
+    "q50": [
+      3.5
+    ],
+    "q90": [
+      6.300000190734863
+    ],
+    "q99": [
+      6.933333396911621
+    ]
+  },
+  "frame_index": {
+    "min": [
+      0
+    ],
+    "max": [
+      210
+    ],
+    "mean": [
+      105.0
+    ],
+    "std": [
+      60.909769331364245
+    ],
+    "count": [
+      1266
+    ],
+    "q01": [
+      2.0
+    ],
+    "q10": [
+      21.0
+    ],
+    "q50": [
+      105.0
+    ],
+    "q90": [
+      189.0
+    ],
+    "q99": [
+      208.0
+    ]
+  },
+  "episode_index": {
+    "min": [
+      0
+    ],
+    "max": [
+      5
+    ],
+    "mean": [
+      2.5
+    ],
+    "std": [
+      1.707825127659933
+    ],
+    "count": [
+      1266
+    ],
+    "q01": [
+      0.0
+    ],
+    "q10": [
+      0.0
+    ],
+    "q50": [
+      2.5
+    ],
+    "q90": [
+      5.0
+    ],
+    "q99": [
+      5.0
+    ]
+  },
+  "index": {
+    "min": [
+      0
+    ],
+    "max": [
+      1265
+    ],
+    "mean": [
+      632.5
+    ],
+    "std": [
+      365.46260638629866
+    ],
+    "count": [
+      1266
+    ],
+    "q01": [
+      12.65
+    ],
+    "q10": [
+      126.5
+    ],
+    "q50": [
+      632.5
+    ],
+    "q90": [
+      1138.5
+    ],
+    "q99": [
+      1252.35
+    ]
+  },
+  "task_index": {
+    "min": [
+      0
+    ],
+    "max": [
+      0
+    ],
+    "mean": [
+      0.0
+    ],
+    "std": [
+      0.0
+    ],
+    "count": [
+      1266
+    ],
+    "q01": [
+      0.0
+    ],
+    "q10": [
+      0.0
+    ],
+    "q50": [
+      0.0
+    ],
+    "q90": [
+      0.0
+    ],
+    "q99": [
+      0.0
+    ]
+  },
+  "observation.images.top": {
+    "min": [
+      [
+        [
+          0.0
+        ]
+      ],
+      [
+        [
+          0.0
+        ]
+      ],
+      [
+        [
+          0.0
+        ]
+      ]
+    ],
+    "max": [
+      [
+        [
+          1.0
+        ]
+      ],
+      [
+        [
+          1.0
+        ]
+      ],
+      [
+        [
+          1.0
+        ]
+      ]
+    ],
+    "mean": [
+      [
+        [
+          0.09851008774910228
+        ]
+      ],
+      [
+        [
+          0.08292840844862344
+        ]
+      ],
+      [
+        [
+          0.07099545462517447
+        ]
+      ]
+    ],
+    "std": [
+      [
+        [
+          0.08389340011198143
+        ]
+      ],
+      [
+        [
+          0.08916203412909407
+        ]
+      ],
+      [
+        [
+          0.07551526948420047
+        ]
+      ]
+    ],
+    "count": [
+      1266
+    ],
+    "q01": [
+      [
+        [
+          0.06666666666666667
+        ]
+      ],
+      [
+        [
+          0.047058823529411764
+        ]
+      ],
+      [
+        [
+          0.0392156862745098
+        ]
+      ]
+    ],
+    "q10": [
+      [
+        [
+          0.06666666666666667
+        ]
+      ],
+      [
+        [
+          0.050980392156862744
+        ]
+      ],
+      [
+        [
+          0.047058823529411764
+        ]
+      ]
+    ],
+    "q50": [
+      [
+        [
+          0.06666666666666667
+        ]
+      ],
+      [
+        [
+          0.050980392156862744
+        ]
+      ],
+      [
+        [
+          0.047058823529411764
+        ]
+      ]
+    ],
+    "q90": [
+      [
+        [
+          0.17647058823529413
+        ]
+      ],
+      [
+        [
+          0.14901960784313725
+        ]
+      ],
+      [
+        [
+          0.12156862745098039
+        ]
+      ]
+    ],
+    "q99": [
+      [
+        [
+          0.5686274509803921
+        ]
+      ],
+      [
+        [
+          0.5725490196078431
+        ]
+      ],
+      [
+        [
+          0.5215686274509804
+        ]
+      ]
+    ]
+  },
+  "observation.images.wrist": {
+    "min": [
+      [
+        [
+          0.0
+        ]
+      ],
+      [
+        [
+          0.0
+        ]
+      ],
+      [
+        [
+          0.0
+        ]
+      ]
+    ],
+    "max": [
+      [
+        [
+          1.0
+        ]
+      ],
+      [
+        [
+          1.0
+        ]
+      ],
+      [
+        [
+          1.0
+        ]
+      ]
+    ],
+    "mean": [
+      [
+        [
+          0.21466016871134228
+        ]
+      ],
+      [
+        [
+          0.19365195997162554
+        ]
+      ],
+      [
+        [
+          0.18467655265444619
+        ]
+      ]
+    ],
+    "std": [
+      [
+        [
+          0.2835314091248394
+        ]
+      ],
+      [
+        [
+          0.280709241715533
+        ]
+      ],
+      [
+        [
+          0.2755891335095306
+        ]
+      ]
+    ],
+    "count": [
+      1266
+    ],
+    "q01": [
+      [
+        [
+          0.058823529411764705
+        ]
+      ],
+      [
+        [
+          0.043137254901960784
+        ]
+      ],
+      [
+        [
+          0.03137254901960784
+        ]
+      ]
+    ],
+    "q10": [
+      [
+        [
+          0.06666666666666667
+        ]
+      ],
+      [
+        [
+          0.050980392156862744
+        ]
+      ],
+      [
+        [
+          0.047058823529411764
+        ]
+      ]
+    ],
+    "q50": [
+      [
+        [
+          0.06666666666666667
+        ]
+      ],
+      [
+        [
+          0.050980392156862744
+        ]
+      ],
+      [
+        [
+          0.047058823529411764
+        ]
+      ]
+    ],
+    "q90": [
+      [
+        [
+          0.8470588235294118
+        ]
+      ],
+      [
+        [
+          0.8313725490196079
+        ]
+      ],
+      [
+        [
+          0.8117647058823529
+        ]
+      ]
+    ],
+    "q99": [
+      [
+        [
+          0.8784313725490196
+        ]
+      ],
+      [
+        [
+          0.8666666666666667
+        ]
+      ],
+      [
+        [
+          0.8509803921568627
+        ]
+      ]
+    ]
+  }
+}

meta/tasks.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1723a05a42ddd3fb8b6f7001dac4ce82065ff713e1a1e3db93a83d51d2b8d7aa
+size 2337

videos/observation.images.top/chunk-000/file-000.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e608a75878c18dbfc3adda16ba8fd262ddca667fd88c9a7ef946b4bd78ca043b
+size 212190

videos/observation.images.wrist/chunk-000/file-000.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:171c3ae69dbd0a2f9bc71889c01702504e726f4ff9102073c89a7bc09998487f
+size 372043